Google slips as OpenAI said to be working on search product: report
seekingalpha.comI'm long GOOG (horizon of 3 years) as 1) their current profits are stratospheric and growing 2) the companies that make money off of AI will already have a strong cloud offering. My thesis is that B2B is where the money is at. B2C is trash unless you can insert ads. B2C subscriptions for something like search isn't even with talking about.
Monetization of AI hasn't even really started yet so I could be tragically wrong. It'll be an interesting ride no matter what.
I asked Kagi search's LLM if cloud could make up for their consumer business tanking, and it pointed out (summarizing, it wrote a half page report with references) their cloud division is only 20% of their current revenue, it lost $5.6B on $13B of revenue in 2021, and that its growth is slowing.
For the same query ("assuming they lose search and advertising google cloud revenue likely to make up the difference in revenue?"), google responded with the following top hit:
> Simply Wall Street - Jul 16, 2023 — I assume that Google's Cloud revenue will grow at 18.5% for the next 5 years, in line with expectations of the industry's growth rate.
Long GOOG (3 years) might make sense though. Word in the valley is they're trying to increase margins by driving costs down. That should improve their multiplier even as they retreat from core markets and their revenue drops.
[edit: I summarized what it said wrong. Ad revenue is 80%. Everything else (including Android licensing and hardware sales) is 20%. It didn't give a percentage for cloud revenue.]
Search subscription seems to be working quite well for Kagi. Source: I use it, and I like it. If you search HN for Kagi you'll find a lot of discourse about it.
I use Kagi too. Buy there's just no universe where selling subscriptions to search is going to generate FAANG level revenues.
As bad as Google Search is I‘m cautious about the prospects of this making a dent in their search monopoly.
Remember when ChatGPT was integrated into Bing? Turns out that did nothing for search market share.
Search market share is all about placement in the browser experience. The most is the browser share. If anyone remembers the browser wars and portal wars that’s what that was all about. Until chrome is dead and Google stops buying the front door on most other alternatives their search engine will dominate. Not because it’s better but because it’s the lowest energy configuration for accessing internet content.
Remember when Bing was integrated into chatGPT? Now I do about half my searches through Bing.
But, it is unclear whether there is room for advertising in chatGPT the way there is room for ads in search. But I suspect there will somehow be a way!
Bing's reputation is such that they'd have to be giving out free money for people to switch.
Besides, GPT was only available on Edge. Again, reputation.
How is OpenAI going to monetize it? The first time you get an ad in your chat is going to really cause some breaking of the fourth wall.
Maybe it would be for paid customers only.
Alternatively, if OpenAI thinks search queries are an important signal, the search engine could be free with no ads for everyone. For example, I was surprised that Amazon pays Prime members $20 for using Amazon Photos. I think they might be using Photos to get more data for AI training.
This can be summarized as a bet on the cost per query going down significantly. Last I checked, it costs OpenAI almost 20x more to run a query vs Google. At that price-point, without serious monetization and even assuming GPU prices go down significantly within the next 5 yrs, it's really more a bet on being to financially withstand the competition.
So Google is really becoming the IBM of this era?
Tech has a way of ensuring relative irrelevance for companies that achieve market dominance.
if it has any whiff or mention of bing, hard pass.
I don't really think technology is the limiting factor in search quality right now. Adding more technology doesn't address the problem.
How about just a business model that isn't awful/predatory?
Perplexity pro for the win.
Good news. Better options are what we need. Using Google search is now pointless. Recently, the page cache feature was removed. Very likely the last excuse to use Google search at all.
Google didn't slip. Google's stock price fluctuated. Which is a normal occurrence. There's a difference between the two, and the headline is really close to click-bait.
There is an entire industry of companies that attach stock dips to whatever negative news they can find for it. Often this is done automatically and the articles published automatically.
Whenever a stock dips, a whole bunch of people want an explanation why, so they google it and read the meaningless article sandwiched between layers of ads.
Fortunately, those ads are often served by GOOGL, so viewing that explanation helps it recover.
Makes sense why we are seeing mass layoffs. It feels like the end of a tech era and companies must rapidly shift priorities
Oh. This seems to imply at one point they thought ChatGPT was something more than the next-gen search engine.
I asked this three months ago: https://news.ycombinator.com/item?id=37975601#37975983
I have been using Kagi paid for the last 2 months and have found it an acceptable drop in replacement for Google. Before that I used Duck Duck Go and found myself using the !g operator about 50% of the time.
If Google dies Linux will be loosing android and investment it provides for Firefox as well.
Hello OpenAI,
Please avoid "personalized" ads.
DuckDuckGo has proven that personalized ads are not necessary for success.
DuckDuckGo injects affiliate links into eBay search results, depends on Bing for its search index and has MS trackers in its app (or at least use to, I stopped caring). You should take a look at Brave Search instead for an independent, privacy-friendly search engine.
Other than the MS trackers, how does any of that impact privacy?
As for the MS thing, they did fix it:
https://techcrunch.com/2022/08/05/duckduckgo-microsoft-track...
and they were never part of the apps. They previously whitelisted microsoft ad and tracking domains in their browser.
I read through the linked article, and although they did seem to address the MS tracking somewhat, it looks like the state of DDG is even worse. It has all sorts of MS tracking stuff baked into the search itself with the whole thing seemingly just based on MS doing a pinky promise not to use the data they're getting.
That's not true - there's no way for to make search histories from DDG searches: https://duckduckgo.com/privacy
> Despite this expansion of DDG’s ability to block Microsoft tracking requests, there are still instances where Microsoft ad scripts are not blocked by DDG’s tools by default — related to processes used by advertisers to track conversions (i.e. to determine whether an ad click actually led to a purchase).
...
> DDG says it wants to go further to protect user privacy around ad conversion tracking — but admits this won’t happen any time soon. In the blog post Weinberg writes that “eventually” it wants to be able to replace the current process for ad conversions checks by migrating to a new architecture for assessing ad effectiveness privately.
From the linked article. Which points us to the ad disclosure[0]:
> If you click on a Microsoft-provided ad, you will be redirected to the advertiser’s landing page through Microsoft Advertising’s platform. At that point, Microsoft Advertising will use your full IP address and user-agent string so that it can properly process the ad click and charge the advertiser.
followed with
> Microsoft Advertising does not associate your ad-click behavior with a user profile. It also does not store or share that information other than for accounting purposes
Hence the pinky promise. I'm good on all of this personally.
[0] https://duckduckgo.com/duckduckgo-help-pages/company/ads-by-...
Brave is not secure or a valid option imo
Care to elaborate?
So far OpenAI has bet that people are simply willing to pay for good services. We'll see if that holds.
Hopefully they'll also allow people to use it in logged-out / memory-less mode.
Duck Duck Go proved search works fine without user profiling, and Google proved personalized search is a clusterfsck.
Logged-out is not really compatible with a paid service, unless you're doing something with anonymous crypto.
There is a big difference between "logged out" and "private mode" - paid VPN services offer privacy by (purportedly) not tracking or logging traffic from their users. They still get paid using traceable payments systems.
This is simply not possible. When I left Google AAR per North American users was $1200/y. I’m betting it’s close to 2k now.
Try turning down that money for 20 bucks a months. Subscriptions will never scale due to one simple fact. When a subscriber pays only one party pays and has a ceiling. When advertisers pay an infinite number of parties can pay per user with no ceiling other than performance.
Ad tech is just getting started.
Surely that can't be right? The US + Canada together have a population of under 380M. Even your low-end ARPU number of $1200 per year would imply North American revenue above $450B, far higher than Google's entire 2023 annual revenue of $300B [0].
(Even correcting for market share isn't enough, as Google Search has over 90% penetration in the US [1].)
[0] https://www.statista.com/statistics/266206/googles-annual-gl...
Definition of user is doing a lot of lifting here, if I remember it was like signed in, 1da, 2+ products. Scope and scale at google made things really different than other places.
They defined users from a business sense as something we wanted to achieve within the product ecosystem -- not just randos using the product occasionally
Looking at how useless "AI-powered Bing" is, I'm not holding my breath for something revolutionary in the space.
Perplexity is actually good though. If I have to guess, OpenAI is going to clone their product.
What makes you say Bing + Copilot is "useless?" It does a lot of stuff.
Yeah, like redirect me to promotions, affiliate links and ads.
Google search is so vulnerable. It's trash for a lot of queries these days, to the point that I'm tempted to switch to Bing. An iterative search tool using a ChatGPT interface on the sidebar that can summarize and cite as it goes would an instant switch.
I'm not sure that being bad is not the same as being vulnerable. Even people who despise Google (like yourself) apparently still use it. Google replaced Altavista et al. because it was miles better than they were. Nothing today is miles better than Google, and there is a lot less competition than there was in 1999. Google is synonymous with search, literally. I can't imagine a scenario in which anyone in my family stops using Google as a search engine; the idea wouldn't occur to them. I've been a Kagi user since the beta, and a DDG user before that, and still don't think Google's going anywhere soon.
My google volume has dropped heavily since chatgpt. I mostly use it to pull up urls for things when I already know what I'm looking for, or if I'm looking for very current information. ChatGPT can use bing which helps a lot with the current information aspect but I haven't found its search performance to be great, it's slow and it really wants to just answer questions briefly in accordance with its alignment. If it was faster and designed more around showcasing the best sites and content (like a search recommender hybrid) while also answering questions directly when possible it would be amazing.
I agree that Google search is very vulnerable. The results are spammy and the current ad volume badly degrades usability for me. In my case, I don’t use Google search much anymore. Previously it was my one stop for finding information online.
Instead of Google search:
I’ve switched almost exclusively to Kagi for vanilla search. It feels like Google in 2015, and the forums/small web filters are great. I do rarely use !g, and usually I'm disappointed.
If I just need a quick question answered I’ll usually use Perplexity.
For coding questions, I mostly use copilot and documentation these days. I barely use search for coding questions, unless I run into a weird edge case, or need to find Github issues for a project. There’s too much link spam to wade through, and copilot is built right into VSCode, so I don’t have to context switch.
I was a big user and early adopter of ChatGPT, but I don’t use it much anymore. Special use models are better for specific use cases (e.g. copilot for coding), and if I want to learn general information about a topic or gather some opinions, I much prefer to search forums and blog posts or watch some YouTube videos (this is the one major bright spot for Alphabet imo). If I do need AI to do something for me, I prefer to use an API or local model. ChatGPT is inferior to the API/local products, especially if you know some Python. This is even more true when you take into consideration copilot (and starts to hint at the compounding opportunity of these tools).
If I want to do a deep dive on something I don’t even bother with most web content these days unless I'm researching a cutting edge topic. The general web is too noisy and inundated with inaccurate and low quality content. Most of the time I either buy a book or directly visit an authoritative source (e.g. the SEC, the courts, Wikipedia, etc.). Less frequently I'll find a niche forum or blog via Kagi's filters, or hit up Google Scholar/Arxiv (another bright spot for Alphabet, but very niche).
The most interesting thing to note here is that I’m paying for almost all of these things, rather than using free Google search, because the Google search user experience has degraded so badly. Kagi, copilot, LLM APIs, books, they all cost money. That doesn’t bode well for the Google search product long term.
> ChatGPT is inferior to the API/local products, especially if you know some Python.
Why is this? GPT-4 outperforms smaller models I can run locally. Is GPT-4 via API better than GPT-4 via webapp?
Yes, for a couple of reasons:
- I control the system prompt to guide the model to do exactly what I want, rather than begging and cajoling ChatGPT. I will often use the API playground for this reason where someone else might use ChatGPT.
- For more complex problems, having the input and output in Python gives me a ton of options for data manipulation, storage, etc.
- Using an API call lets me daisy chain different models and other APIs/tools together. ChatGPT can sort of do this via plugins, but I haven't found it to be a great experience.
GPT-4 is definitely better than the local models by itself. It's not a clear winner when you start to get into fine-tuning/ensemble workflows. It's also less attractive in situations where I want to chew through a ton of tokens for a personal project and GPT-4 would get expensive.
> daisy chain different models and other APIs/tools together
> ensemble workflows
This is intriguing, can you explain in more detail what this looks like?
Look up langchain. You can have most models format output in such a way as to facilitate programmatic execution based on their contents, and you can let the model just run that way with some checks to keep it from losing the plot.
I feel the same. I keep using other search engines to get some useful response or content. I think the AI generated content or promoted links changed the results for the worse these days...
I welcome the completion but surely Google could clone that. The real bet is that they won’t have the courage to do it.
Courage? What in the world does that even mean? How would it take any more “courage” than it took to launch a product that sucks people’s entire website into the search result, guaranteeing an outcry of epic proportion?
FWIW, I tried switching to Bing and it was equal or worse trash. I can't quantify it but I had my default search set to Bing for a few months and I routinely would search, find nothing relevant, then head to Google.
Google sucks but it still seems better than the other bigger players at least. Not a fanboy. I'd like to see something actually useful replace them all but that was my experience.
Sundar Pichai is more of a figure head rather than somebody “leading” the tech giant. You could put any MBA school trained flunkie in his position and the result will be the same.
Glad Google is getting stormed by the competition.
> You could put any MBA school trained flunkie in his position and the result will be the same.
Most managers (at all levels) could be replaced with a Magic 8 Ball, and you would get similar results. Although, in some cases, consistency would improve.
Here is the thing about Sundar, emphasis mine: https://share.zight.com/7KunAo4K
I always wonder why they keep him around when it's clear as day to anyone paying attention he's... not the best.. but.. that graph somewhat speaks for itself.
"Stormed by the competition" by what metric? And no, Substack blogs and HN comments don't count.
Don't forget, power user recommendations got google/gmail off the ground.
I can count on one hand how many legacy brave searches I've done this week because you have to wait 5 seconds for the new summary to load. When I do use traditional search instead of chatbot it usually comes out of respect for the free perplexity.ai chatbot I'm using, I feel bad asking it a legacy question like the acronym for something when i can just highlight>right click> search in new tab.
I already stopped using Google for most programming questions, I ask ChatGPT instead. Recency of results is an issue though, but I think if OpenAI nails a way to e.g. search for products or services they will be able to disrupt Google easily. Most content you find on Google is SEO trash, an LLM agent should be able to cut through that for me and save quite a lot of time.
This is one of the things I like about Cursor (https://cursor.sh/, a fork of VS Code that adds AI features).
It uses GPT (I believe they're running their own fine-tuned version), and allows you to '@' items like your files, folders, and (most impressively) any documentation.
So, if I'm working on something that uses React Hook Form, I can paste the URL to the RHF documentation. Cursor will index it (presumably creating embeddings for each page), and then I just '@' the RHF documentation in the chat and it'll find the relevant pages and include it them the context for GPT to answer.
It's very useful. I already subscribe to ChatGPT Plus but Cursor adds something special that I'm more than happy to subscribe to Cursor too.
How did you find out about Cursor?
OpenAI needs to be fast (few ms), and be able to show sources (links)
Does it, though? I do wonder if we just have set up a false expectation of how search ought to work. I want answers to problems I have, 20-30 seconds for a good quality answer will very, very often be vastly preferred over 200ms and results I'll have to frequently delve deeper into myself, and I am not adding any less than 30 seconds in most cases.
I'm happy to use a !bang to indicate what kind of response I want. I'm not the average user but people understand the difference between asking your roommate who's right there and asking a professional about something, so this doesn't seem like much of a stretch.
I tried to use chatgpt to search for many things, but it's just too slow for the amount I use google. Adding the lengthy prompts does not help.
Performance is a feature
OpenAI is already far faster than Google. Sure, Google returns your search results very quickly, but it then takes me forever to sift through the results and tweak my search query until I _maybe_ find an answer to my question, and I'm finding more and more that no matter how long I take, Google can't give me the answer at all.
GPT-4 takes maybe 15-30 seconds to spit out a precise, detailed answer to exactly what I asked for, and it's usually the correct answer, or at least it's close enough. It would have to get a _lot_ slower to make me go back to Google.
It also needs to be profitable, obviously their free tier is a money sink but I think it's currently unclear whether even their paid tiers are net-positive.
Even if they provide a better service, if it's only feasible to run when subsidised by the AI hype-wave VC money firehose then it's not going to last.
There is a trade off between processing time and quality of result's that pays off if the first result is the answer sought.
Very true.. I resort to chatgpt if things are not clear enough, when I need a broader picture, or when I want to have a discussion about the matter.
Note the 'resort'. Huge opportunity here indeed
This is about infra rollout. The thing for Google is they can’t just slap Bard on the home page because it would be flattened in 10 seconds unless they massively invest in GPU resource.
So what’s Google’s play here? Charge money to deliver ads to people?
A startup otoh can start small and grow as revenues are received.
The play? Advice people on what to buy. Have a conversation about the user's needs. Discover latent needs.
They have the opportunity to do automated solution sales.
"Bard Qualified Leads", they can charge a commission instead of CPM. Genius!
I recommend Phind if you want to sources.
Interesting take! Quick results (not as quick as google) + a summary / answer to the question.
I just would like to be able to disable: "To improve our service, we may anonymously log your search queries and their associated results"
I know google does the same, but there's more control there, and a large company is less likely to look into my specific queries.
It needs to not advertise.
I had a Vue question and Google was utterly useless and ChatGPT was giving only answers about the Options (legacy) API. I needed answers for the Composition API. So I told ChatGPT this is syntax for Options <code> and this is syntax for Composition <code> please restrict your answers to Composition examples and it worked.
Sounds like you're on the free version of Chat GPT. What was your prompt?
This was after going a few rounds of prompting in a more search-engine style
for pinia, the composition api syntax is in the form: "defineStore('myStore', () => {})" and options api of the form: "defineStore('myStore', {})", using this definition, please show an example of using pinia and reactive() with the composition api instead of the options api results previously shown. thanks
I've been using phind almost exclusively because it cites sources in the sidebar and it's free if you use their model. If I need code generated, ChatGPT will generally do a better job, though.
> Most content you find on Google is SEO trash, an LLM agent should be able to cut through that for me and save quite a lot of time.
I don't believe this will last.